'Face detection and recognition' I want to display test image and recognised image and if match not found i want display match not found message

DBPath='D:\Matlab_prblms\face_database';
D=dir(DBPath)

[mFiles t]=size(D)
mSamples= mFiles-2;
A=zeros(size(image,1),mSamples);
size(A)

for i= 3 : mFiles
    D(i).name
    image=imread(strcat(DBPath,'\',D(i).name));
    [x, y, z]= size(image)
    if(z==3)
          image=rgb2gray(image);
    end
    image=imresize(image,[64,64]);
    image = image(:);
    A(:,i-2)=double(image);
end
%At=A.';s
M=mean(A');
phi=(A-M');
C=phi'*phi;
k=10;
[V D]= eigs(C,k);
size(phi)
size(V)
U=phi*V;
size(U)

for i=1 :k
length=norm(U(:,i));
U(:,i)=U(:,i)./length;
end

wi=zeros(k,mSamples);
for i=1 : mSamples
    wi(:,i)= U'*A(:,i);
end

testImage = cap_img();
[
% testImage=imread('D:\Matlab_prblms\face_database\17.tif');
testImage = rgb2gray(testImage);

testImage=imresize(testImage,[64 ,64]);
testImage=testImage(:);

t=U'*double(testImage);

mindist=999999;
index=-1;
for i=1 : mSamples

      distVal = (wi(:,i) - t).*(wi(:,i) - t);
      if(mindist > distVal)
          mindist = distVal;
          index=i;

      end
  end
  for i = 1 : mSamples

      q = t(:,i);
      temp = ( norm( t - q ) )^2;
      mindist= [mindist temp];
  end
  index

[Carl]

Hi Santhosh, can you clarify the problem for us?

1. Is there a specific person's face you are trying to identify in a picture? Or are you looking for any face in general?
2. What are the "test" and "recognized" images for?
3. Is there a specific part of this workflow, or your code, that you are having trouble with?

[Santhosh Nayak]

Im recognizing face from image i captured with web cam and comparing with images stored in the folder and im displaying the matching image also but the problem is in between if matching image not found i want to display match not found how i can do that can u help here i copied my code have a look

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

clear all;
close all;
DBPath='D:\Matlab_prblms\face_database';
D=dir(DBPath)

[mFiles t]=size(D)
mSamples= mFiles-2;
A=zeros(4096,mSamples);
size(A)
name_cell={D.name};

for i= 3 : mFiles
    D(i).name
    image=imread(strcat(DBPath,'\',D(i).name));
       [x, y, z]= size(image)
    if(z==3)
          image=rgb2gray(image);
    end
    image=imresize(image,[64,64]);
    image = image(:);
    A(:,i-2)=double(image);

end
%At=A.';s
M=mean(A');
phi=(A-M');
C=phi'*phi;
k=10;
[V D]= eigs(C,k);
size(phi)
size(V)
U=phi*V;
size(U)

for i=1 :k
length=norm(U(:,i));
U(:,i)=U(:,i)./length;
end

wi=zeros(k,mSamples);
for i=1 : mSamples
    wi(:,i)= U'*A(:,i);
end

testImage = cap_img();   % this is were im passing captured face image for comparing

img=testImage;
% testImage=imread('D:\Matlab_prblms\face_database\17.tif');
testImage = rgb2gray(testImage);

testImage=imresize(testImage,[64 ,64]);
testImage=testImage(:);

t=U'*double(testImage);
% imshow(t)
mindist=999999;
index=-1;
for i=1 : mSamples

      distVal = (wi(:,i) - t).*(wi(:,i) - t);
      if(mindist > distVal)
          mindist = distVal;
          index=i;

      end
  end
  name_cell{index+2}
  matchedimage=imread(strcat(DBPath,'\',name_cell{index+2}));
  % imshow(matchedimage)
  figure
  subplot(2,2,1),imshow(img),title('Input image');
  subplot(2,2,2),imshow(matchedimage),title('Match Image');

[Carl]

I'm not sure I completely understand your code. I do not know what method you are using to compare the webcam picture with your stored pictures. Are you checking for a matching face based on the mindist value? In that case would you be able to threshold on some value to determine if the image corresponding with mindist is a "match" or not?

[Santhosh Nayak]

yes Im using eigenvectors mindist so in that case how can display message if matching image in folder not found

[Carl]

Are you asking how to display the message, or how to identify when there is no match? For displaying a message, you can use something simple like the "disp" function. For identifying if there is no match, that would depend on how you interpret that mindist value.

Either way, I would strongly suggest that you follow the link posted by Image Analyst below. As they mentioned, it contains a much more robust approach to your original problem.

Answer 1

[Image Analyst]

I doubt any robust face recognition code is going to be only like 100 lines of code. You may want to look at how Brett, of The Mathworks, did it here: http://www.mathworks.com/matlabcentral/fileexchange/49914-streaming-face-detection--training--recognition

[Santhosh Nayak]

In that link they are making use of surf and fast but I'm using PCA eigenvector, and they have created montages. If tried to implement that, I have to re-code everything is there any easy method to find min distance and compare suggest if any.

I tried ismember but that isn't working. Is there any other way that I can use?

Thank you

[Image Analyst]

It seems you're committed to getting your method working and don't want to make major changes to your existing code.

To find minimum distances I always use the Pythagorean theorem, sqrt(), to get distances and then use min() to find the min of the distances, OR use pdist2() to find distances of every point to every other point and then use min(). Whether one of those two methods will help you, I don't know. That's about all I can say.